3 November 2009 | Team Tamar

When will the web know what I want?

So you need to find the answer to a question but the question you’re asking doesn’t have a ‘fixed’ answer. How does the web know what you are actually looking for and how does it know what it has will actually match it? Do we need a ‘new’ search engine or can we improve on existing services?

Typical questions with accepted ‘fixed’ answers.

‘What is 2+2?’ – The only answer is 4 (unless using a strange form of maths).

‘What is the capital of England?’ – The only answer is London and there is no other response.

There is only one answer that is correct and as such it is relatively easy to find an answer. What happens then, when you ask a subjective question or one based on a specific context?

Where is a good place to eat?’ or ‘Is the film 9 any good?’ – If you Google the first, it points you to Yahoo! Answers as the top result and if you Google the second, it comes up with things from ‘negative scanner’ discussions about film negative types on Flickr to ‘the 9 sassiest black women in film’.

Did anything good come out of it?

Now this isn’t completely without merit, there are some sites listed in the results that would match the queries I have asked more, but don’t appear near the top. The first search about places to eat suggests an article by The Observer about the ‘100 best places to eat this summer’. It’s good, but it’s not the one. It’s four years out of date and while many of the places in the list are probably still open, it is not specific enough to my query. In other words it didn’t understand me or relate it to my current situation.

The second has a brief article about the film (the right film) and even has a video review which isn’t bad but it took some extra time to find. This is one person’s opinion and while it may be accurate and reflect my own, I was looking for a more authoritative review or one that averages ratings like IMDB.

The point is that I want Google et al. to know exactly what I want.

Improve the results = kill all humans

Now I could do things to increase the likelihood of the search hitting the desired outcome such as adding speech marks around specific phrases or adding in my location but the important difference I could make would be to rephrase so as not to be asking a question, to ‘de-humanise’ the search.

So ‘Where is a good place to eat’ could become ‘good places to eat in Chiswick’ (which sounds like it’s only part of a sentence) and ‘Is the film 9 any good’ becomes ‘9 film review’ (which sounds like nothing anyone would say in conversation.

The first result for ‘good places to eat in Chiswick’ provides a list of restaurants in Chiswick from TopTable.com, ordered by rating and giving details of the type of cuisine and average cost. Perfect. Even without the ‘good’ search term I am presented with a page about Chiswick with emphasis on cafes and restaurants.

The first result for ‘9 film review’ is ‘9 – Film Review’ from Hollywoodreporter.com which contains a very balanced review that was written yesterday, exactly what I wanted.

Why can’t I just ask a question?

So why, when I ask a question, do I not get the answers I want and only when I talk like a machine do I get the exact matches? Well the answer may lay in talking like a machine… to a machine. Google understands the relationship between keywords and how to rank pages to make sure the pages at the top of the results are the most authoritative, but it doesn’t understand context.

“Right now, Google is really good with keywords and that’s a limitation we think the search engine should be able to overcome with time,”

“People should be able to ask questions and we should understand their meaning, or they should be able to talk about things at a conceptual level. We see a lot of concept-based questions — not about what words will appear on the page but more like ‘what is this about?’. A lot of people will turn to things like the semantic web as a possible answer to that.”

Marissa Mayer (Google Vice President of Search Products & User Experience) from an interview with IDG News Service in October 2007.

What does this mean then?

Generally, the current level of search is fine. People who know how to search for things on the internet tend to use Google and tend to do a few variations / combinations of search terms to hone in on a result matching what they want. While time consuming, the method of just typing in keyword combinations is quicker than asking a whole question.

So how can Google supplement its dominating search algorithms with semantic understanding? With more and more information about ‘you’ and your browsing habits spanning more than one website (Facebook connect for example), is it possible this could reach further and be more tightly integrated into the process of searching? There is already location-aware search.

If this is the case then where is the line drawn between accurate and contextually accurate search results, and privacy of our information?

Would you let Google know everything about you in order to get perfect search results?

Team Tamar